1. Field of the Invention
The present invention is directed to scanning probe microscopes (SPMs), and more particularly, to alternate methods of accounting for the shape of the probe tip in the acquired image, the methods being particularly adapted for measuring critical dimensions (CD) and features of semiconductor wafers, data recording media, and related.
2. Description of Related Art
Several known probe-based instruments monitor the interaction between a cantilever-based probe and a sample to obtain information concerning one or more characteristics of the sample. For example, SPMs, such as the atomic force microscope (AFM), are devices which typically use a sharp tip and low forces to characterize the surface of a sample down to atomic dimensions. More particularly, SPMs monitor the interaction between the sample and the tip (where the tip is typically mounted on the cantilever of the probe). By providing relative scanning movement between the tip and the sample, surface characteristic data can be acquired over a particular region of the sample, and a corresponding map of the sample can be generated.
The atomic force microscope (AFM) is a very popular type of SPM. The probe of the typical AFM includes a very small cantilever which is fixed to a support at its base and which has a sharp probe tip extending from the opposite, free end. The probe tip is brought very near to or into contact with a surface of a sample to be examined, and the deflection of the cantilever in response to the probe tip's interaction with the sample is measured with an extremely sensitive deflection detector, often an optical lever system such as described in Hansma et al. U.S. Pat. No. RE 34,489, or some other deflection detector such as strain gauges, capacitance sensors, etc. The probe is scanned over a surface using a high resolution three axis scanner acting on the sample support and/or the probe. The instrument is thus capable of creating relative motion between the probe and the sample while measuring the topography or some other surface property of the sample as described, e.g., in Hansma et al. U.S. Pat. No. RE 34,489; Elings et al. U.S. Pat. No. 5,226,801; and Elings et al. U.S. Pat. No. 5,412,980.
AFMs may be designed to operate in a variety of modes, including contact mode and oscillating mode. In contact mode operation, the microscope typically scans the tip across the surface of the sample while keeping the force of the tip on the surface of the sample generally constant. This effect is accomplished by moving either the sample or the probe assembly vertically to the surface of the sample in response to sensed deflection of the cantilever as the probe is scanned horizontally across the surface. In this way, the data associated with this vertical motion can be stored and then used to construct an image of the sample surface corresponding to the sample characteristic being measured, e.g., surface topography. Alternatively, some AFMs can at least selectively operate in an oscillation mode of operation such as TappingMode™. (TappingMode™ is a trademark of the present assignee.) In oscillation mode, the tip is oscillated at or near a resonant frequency of the cantilever of the probe. The amplitude or phase of this oscillation is kept constant during scanning using feedback signals, which are generated in response to tip-sample interaction. As in contact mode, these feedback signals are then collected, stored, and used as data to characterize the sample.
Regardless of their mode of operation, AFMs can obtain resolution down to the atomic level on a wide variety of insulating or conductive surfaces in air, liquid or vacuum by using piezoelectric scanners, optical lever deflection detectors, and very small cantilevers fabricated using photolithographic techniques. Because of their resolution and versatility, AFMs are important measurement devices in many diverse fields ranging from semiconductor manufacturing to biological research.
Notwithstanding the fact that scanning probe microscopes are high resolution instruments, the ultimate resolution of the data obtained by such probe-based instruments is limited by the physical characteristics of the tip of the probe itself. More particularly, there are limitations as to how small, and sharp, the tip can be made. In view of this, the tip shape is reflected in the acquired data, a problem that is exacerbated by the fact that AFMs often image very small (e.g., Angstrom scale) features. As a result, an error in the acquired data results and the corresponding accuracy of the surface image is significantly compromised. Hereinafter, the acquired SPM image will periodically be called the “dilated” image.
For some applications, this limitation may be negligible. However, for many applications, the degree of accuracy required to resolve the features of the sample surface is significantly greater, such that tip shape error is unacceptable. For instance, in the semiconductor fabrication industry, imaging features such as lines, trenches and vias with single nanometer accuracy is desired. These features may have dimensions in the range of about 90 nm, and are continually getting smaller. With typical tip widths in the range of about 70 nm, the tip shape clearly introduces significant error in the data and must be removed to accurately image the sample surface.
Moreover, the aforementioned problems can be exacerbated by the fact that complex sample surface topologies require a commensurate increase in tip shape complexity to image such surfaces. For example, samples may include undercut regions where a particular x,y scan position may have multiple “Z” height values (see region “U” in FIG. 1 , discussed in further detail below). Again, this is common in the semiconductor fabrication industry, and thus tips have been developed to allow imaging of such complex topographies. However, with the increase in tip shape complexity, there typically is a corresponding increase in error in the AFM data.
Two types of known tip shapes are illustrated in FIGS. 1 and 2. Note that probe tips, such as the CD tip, shown in FIG. 2, typically will not have the smooth symmetrical shapes illustrated in the figures. These tip shapes are merely presented as such to highlight the concepts and features of the preferred embodiment. In FIG. 1, a probe tip 10 of a traditional scanning probe microscope includes a parabolic, or other pointed shape that is relatively easy to characterize. Tip 10 includes a shaft 12 and a distal end 14 that although sharp is typically at least slightly rounded at its active surface 15. During a scan (operating in an oscillating mode, for instance), tip 10 interacts with a sample surface 16 to image characteristics of that surface. Tip-sample interaction is controlled, and data is collected, via a control system (not shown) as described previously. The collected data, in turn, may be plotted to image the sample surface. Importantly, this acquired image may not accurately reflect sample surface characteristics due to, among other things, the error introduced by the shape of the pointed tip.
In addition to introducing at least some tip shape error in the acquired data, probe tip 10 is unable to image certain surfaces. In particular, although suitable for many applications, based on its shape probe tip 10 is simply unable to accurately depict vertical sidewalls and undercut regions (which often exist in semiconductor fabrication, for example) in the corresponding sample surface topography. Notably, this is due to limitations in both the tip shape and the algorithms used to control tip position.
To be able to image surface features such as vertical sidewalls and undercut regions, AFMs having more complex probe tips have been developed. In one such instrument, shown in FIG. 2, an AFM employs active X-Z control to follow complex surface topography using a probe tip 20 having a shaft 22 and a distal end 24 including left and right protuberances 26, 28, respectively, in the scan (for example X) direction. By dithering the tip in the scan direction, protuberances 26, 28 are caused to interact with surface features such as vertical sidewalls. As a result, what before caused “shaded regions” (i.e., regions of no tip-sample contact such as undercut region “U” illustrated in FIGS. 1 and 2) in the acquired AFM data, now yields at least some data based on tip-sample contact. However, with this increase in flexibility of the types of samples that can be imaged, correcting and reconstructing the image data becomes increasingly difficult.
Overall, whether employing simple or complex probe tip shapes, the problem of the shape of the tip being convolved in the AFM data has been known and appreciated in the art. Although solutions have been attempted with some success for conventional AFMs, extracting tip shape errors from CD AFM data has been an inexact process. Moreover, as features become smaller, and because the tip is at least somewhat limited in just how small it can be made, the convolution of the tip in the image data becomes more substantial, and thus it is becoming increasingly important that the tip shape be removed for accurate measurements.
In another known and widely used technique, particularly applicable to the above-described CD probe shown in FIG. 2, rather than applying shape “deconvolution” of the image to compensate for the effect of dilation of the image, a simple subtraction of the tip-width in the scan direction can provide improved reconstructed images and critical dimension measurements.
For this technique to provide a useful correction, the width of the CD tip must be determined to a high degree of accuracy. The way in which this is typically accomplished is by scanning a silicon nanoedge with, for example, the boot shaped CD tip shown in FIG. 2. Because the dimensions of the nanoedge are known or at least very closely approximated, the width of the tip can be extracted from the image data. This scan of a silicon nanoedge is illustrated in FIG. 3A. In particular, a CD tip (for example, 20 in FIG. 2) is scanned from left to right over an improved silicon nanoedge (ISNE) 31 so as to produce an image data profile 30. In this method, the width of the tip is calculated according to,Wtip=L−(W1+W2)  Equation 1where “L” is the total width of the acquired image a vertical distance “D” (defined below) from the plateau. W1 and W2 are defined as follows,W1=(D−r)tan α+r  Equation 2andW2=(D−r)tan β+r.  Equation 3
In these equations, “D” is the distance from the plateau “P” of the scanned image used for measuring the angles α and β, as illustrated in FIG. 3. For example, this value may be approximately 800 angstroms. In addition, “r” is the radius of the vertex of the ISNE, estimated by SEM, TEM and/or sharp tip SPM analysis of the nanoedge, and is approximately 75 angstroms. The angles α and β are the angles computed from the left and right side slopes, respectively, of the previous tip calibration analysis. Computing the tip width in this fashion, this prior art method can be used to subtract off that width from the image data generated during a scan to approximately correct for the error in the image data. Although providing a correction, this method has significant drawbacks.
First, by simply subtracting the tip width from the image data, it is assumed that the tip-sample contact is being made at a particular point, for example, at the vertical tangent of the protuberances of the boot shaped or CD tip (i.e., at point 29, FIG. 2). However, as the tip scans along a particular topography, the contact point of the tip on the sample translates along the surface of the tip and thus the effective width of the tip at the contact point changes. As a result, a single-valued tip width subtraction is inexact. By simply subtracting off a single value tip-width, an error remains in the reconstructed image as each feature of a unique tip shape cannot be fully accounted for in correcting AFM image data. These errors are directed to inaccuracies in the image of the sample surface shape for both topology and CD width measurements at a particular height. Another significant drawback is that the width defined in Equations 2 and 3 set forth above are merely estimates for the actual tip width. As the samples to be imaged continue to demand greater resolution, these equations will become inadequate even for those applications where tip-width correction provides an acceptable correction.
In short, for the applications contemplated by the present invention, no known technique sufficiently accounts for the tip shape when reconstructing CD AFM image data.
For reconstructing non-reentrant, relatively simple topologies, the methods using local slope-matching between the acquired image data and the tip profile have been attempted. A drawback of slope-matching, or “slope-based,” reconstruction methods is that Legendre transforms used in the analysis, which require numerical derivatives of the data, can be highly sensitive to noise in the original image data. A “smoothing” technique may be implemented to reduce the noise enough to allow reconstruction, but such smoothing typically eliminates sharp features, which are often critical to accurate reconstruction.
A known method to reduce the negative effects of noise is use of a median filter in pre-processing image data prior to slope-based reconstruction of the image data. However, due to inherent limitations, the use of median filters alone does not solve the problem of noise amplification and artifact generation in the reconstructed image. Certain known techniques to smooth or reduce noise can also eliminate crucial features in an image, thereby causing false image reconstruction.
In view of the above drawbacks with known methods of smoothing and pre-filtering original image data, an improved method is desired to reduce noise and artifacts prior to image reconstruction. In addition, alternative methods of image reconstruction particularly adapted for reconstructing complex surfaces using complex probes were also desired.